ENTSO-E AISBL • Avenue de Cortenbergh 100 • 1000 Brussels • Belgium • Tel + 32 2 741 09 50 • Fax + 32 2 741 09 51 • info@entsoe.eu • www. entsoe.eu

Solar Eclipse 2015

- Impact Analysis -

Report prepared by

Regional Group Continental Europe and

Synchronous Area Great Britain

19 February 2015

Solar Eclipse 2015

2/21

Contents

Contents ........................................................................................................................................ 2

Analysis for the Continental Europe Synchronous Area ................................................................ 3

1. Introduction ............................................................................................................................ 3

2. Methodology .......................................................................................................................... 3

3. Input Data .............................................................................................................................. 4

4. Results ................................................................................................................................... 7

5. Next Steps ........................................................................................................................... 10

Analysis for the Great Britain Synchronous Area ......................................................................... 12

1. Summary .............................................................................................................................. 12

2. Model for Eclipse Effect ........................................................................................................ 12

3. Model for the Residual Demand ........................................................................................... 12

4. Demand Effects .................................................................................................................... 14

5. Model for Residual Demand ................................................................................................. 14

6. References ........................................................................................................................... 15

Annexes…………………………………………………………………………………………………… 15

Solar Eclipse 2015

3/21

Analysis for the Continental Europe Synchronous Area

1. Introduction

On 20 March 2015 a solar eclipse will pass over the Atlantic Ocean between 07:40 and 11:50 UTC (08:40-

12:50 CET) and the eclipse will be visible across Europe. The reduction in solar radiation will directly affect

the output of the photovoltaics (PV) and for the first time this is expected to have a relevant impact on the

secure operation of the European power system. In the synchronous area of Continental Europe and the

synchronous area of Great Britain preliminary studies to evaluate the impact of the solar eclipse and possible

countermeasures to be taken by the TSOs have been performed.

In 2015 the installed capacity on PV in the synchronous

region of Continental Europe is expected to reach 90 GW

and the eclipse may potentially cause a reduction of the

PV infeed by more than 30 GW during clear sky

conditions. This situation will pose a serious challenge to

the regulating capability of the interconnected power

system in terms of available regulation capacity,

regulation speed and geographical location of reserves.

Although a solar eclipse is perfectly predictable the

transformation from solar radiation to electric power is

associated with uncertainties which call for a careful

coordination throughout the entire interconnected power

system of Continental Europe including adjacent power

systems.

This report addresses in chapters two till five the

following issues for Continental Europe related to the

solar eclipse

Estimation of the installed PV capacity on March

20 2015 per country

Estimation of the PV infeed on March 20 by combining capacity and coincidence factors for each

country with radiation data with and without the solar eclipse

Inventory of potential mitigation measures

In chapters six till ten an analysis of the situation in Great Britain performed by National Grid addresses the

following issues:

Model for Eclipse effect

Model for the Residual Demand

2. Methodology

The assessment is based on the following input per country

a) Installed PV capacity (Pinstalled)

b) Coincidence factor during the hour of year with maximum radiation (75% assumed)

c) Maximum radiation factor for the year during clear sky conditions (average) (Imax)

d) Radiation factors for each hour of March 20 during clear sky conditions (average) (I(t))

Source: http://astro.ukho.gov.uk/eclipse/0112015/

Solar Eclipse 2015

4/21

e) Begin of eclipse, end of eclipse and maximum solar obscuration (average).

The PV infeed during clear sky conditions is then calculated by combining a) to d).

Pclear sky(t) = Pinstalled*0.75* I(t)/Imax

Similarly, the PV infeed considering the solar eclipse is calculated by correcting the normal infeed by e).

Peclipse(t) = Pclear sky(t) * [1 - obscuration factor(t)]

The formula for the obscuration factor is shown in annex 1.

The following assumptions should be mentioned

- Proportional relation between radiation and PV infeed.

- Average data for each country

No change of system load is considered; the study focuses only on PV variation due to solar eclipse.

3. Input Data

In order to calculate the installed PV capacity on March 2015 per country a linear extrapolation of the capacity

for 2012 and 2013 [1] is assumed as a default rule. This estimation was improved by TSOs reviewing and

correcting where necessary. Table 1 shows the development in installed PV capacity (MW) per country.

Solar Eclipse 2015

5/21

End of 2012 End of 2013 March 20, 2015

Albania n/a n/a n/a

Austria 363 613 917

Belgium 2,768 2,983 3,245

Bosnia 0 0 0

Bulgaria 1,010 1,020 1,032

Croatia 0 20 44

Czech Republic 2,087 2,175 *2,100

Denmark 332 548 *600

France 4,060 4,673 5,419

Germany 32,411 35,715 39,734

Greece 1,536 2,579 3,848

Hungary 12 22 *55

Italy 16,479 17,928 19,691

Luxembourg 30 30 30

Macedonia n/a n/a n/a

Montenegro n/a n/a n/a

Netherlands 360 *780 *1,180

Poland 7 7 7

Portugal 242 278 322

Romania 51 1,151 2,489

Serbia n/a n/a *6

Slovakia 523 524 525

Slovenia 201 212 225

Spain *6,320 *6,722 *6,739

Switzerland 437 737 1,102

Turkey 12 18 25

Ukraine West n/a n/a n/a

Total 69,241 78,735 89,335

Table 1 Expected installed capacity on photovoltaic in Continental Europe on March 2015 in MW. The estimate is based on a

linear extrapolation of data for 2012 and 2013 from [1].*marks estimates corrected by request of TSOs. The data of

Spain includes 2.3 GW of concentrated solar thermal power.

The eclipse start and end time and the maximum obscuration per country are derived from [2]. From this

data source 181 locations are available in the region of Continental Europe. For each country the available

location nearest to the geographical centre has been chosen. See Table 2.

Solar Eclipse 2015

6/21

Start eclipse

UTC

End eclipse

UTC

Max

obscuration

Location

Austria 08:31 10:52 62% Klagenfurt

Belgium 08:27 10:45 80% Brussels

Bosnia 08:35 10:54 51% Sarajevo

Bulgaria 08:44 10:59 40% Plovdiv

Croatia 08:33 10:53 58% Zagreb

Czech Republic 08:36 10:57 69% Prague

Denmark 08:40 10:58 83% Arhus

France 08:20 10:38 77% Orleans

Germany 08:33 10:52 76% Kassel

Greece 08:38 10:51 37% Larisa

Hungary 08:39 10:59 58% Budapest

Italy 08:24 10:44 59% Florence

Luxembourg 08:27 10:46 76% Luxembourg

Netherlands 08:30 10:48 80% Nijmegen

Poland 08:45 11:05 67% Lodz

Portugal 08:02 10:12 70% Coimbra

Romania 08:48 11:05 47% Brasov

Serbia 08:39 10:58 51% Belgrade

Slovakia 08:41 11:01 61% Banska Bystrica

Slovenia 08:31 10:52 60% Ljubljana

Spain 08:05 10:18 67% Madrid

Switzerland 08:24 10:44 70% Berne

Turkey 09:01 11:03 25% Ankara

Table 2 Duration and maximum solar obscuration assumed for each country on March 20, 2015. Data are calculated for the

specific location for each country from [2].

Annex 2 contains detailed information on each of the chosen locations.

The radiation factors are provided by RTE [3] and are shown in Table 3.

The radiation factors are available with hourly resolution whereas the obscuration data are calculated with

one minute resolution.

Solar Eclipse 2015

7/21

Max hour 8:00 UTC

9:00 CET

9:00 UTC

10:00 CET

10:00 UTC

11:00 CET

11:00 UTC

12:00 CET

Austria 91% 51% 68% 77% 81%

Belgium 90% 36% 55% 69% 76%

Bosnia 92% 57% 72% 81% 83%

Bulgaria* 90% 65% 76% 81% 80%

Croatia 92% 57% 72% 81% 83%

Czech Republic 91% 51% 66% 75% 78%

Denmark 88% 36% 53% 64% 69%

France 92% 34% 56% 71% 79%

Germany 90% 43% 60% 71% 76%

Greece 91% 69% 81% 86% 86%

Hungary 91% 56% 71% 79% 81%

Italy 92% 53% 70% 81% 84%

Luxembourg* 90% 36% 55% 69% 76%

Netherlands 90% 35% 54% 67% 74%

Poland 90% 49% 64% 73% 75%

Portugal 92% 23% 49% 69% 81%

Romania 90% 65% 76% 81% 80%

Serbia 90% 62% 75% 82% 82%

Slovakia 91% 55% 70% 78% 80%

Slovenia 92% 54% 70% 79% 82%

Spain 92% 29% 53% 71% 82%

Switzerland 91% 45% 63% 76% 81%

Turkey* 90% 65% 76% 81% 80%

Table 3 Clear sky radiation factors for hour of year with maximum radiation and for the relevant hours of March 20 without solar

eclipse. [3]* marks countries where data is not available. Instead, the nearest country is chosen.

4. Results

Figure 1 shows the total impact of the solar eclipse on the continental European power system. Compared

to clear sky conditions the PV output may drop by 34 GW at 9:41 (UTC) or 10:41 (CET).

Figure 1 Comparison of expected infeed from solar on March 20 during clear sky conditions with and without solar eclipse.

Solar Eclipse 2015

8/21

Figure 2 shows that the minute-to-minute power gradient may exceed -400 MW/minute and

+700 MW/minute. Fortunately, the highest gradient occurs when the PV infeed returns (requiring a

reduction in the load following reserves). Note that a gradient exceeding -400 MW/minute persists for half

an hour.

Figure 2 Expected power gradient in MW per minute from solar power on March 20 during clear sky conditions with and without

solar eclipse.

The same result can be seen in Figure 3 the power gradient over 5 minutes.

Figure 3 Expected power gradient in MW per 5 minutes from solar power on March 20 during clear sky conditions with and

without solar eclipse.

Solar Eclipse 2015

9/21

Table 4 shows that 50% of the infeed reduction is expected in Germany and that Italy amounts for 21%. This

indicates that the risk of line overloading shall especially focus on this region. In addition this table shows

the expected maximum power gradient for each country and Figure 4 the corresponding power gradient for

the five countries mostly affected by the eclipse.

Installed

capacity MW

Reduction

09:41 (UTC)

Reduction

share

Min gradient

MW/min

Max gradient

MW/min

Albania n/a

Austria 917 346 1% -6 7

Belgium 3,245 1,360 4% -21 31

Bosnia 0 0 0% 0 0

Bulgaria 1,032 258 1% -5 6

Croatia 44 16 0% 0 0

Czech Republic 2,100 838 3% -14 18

Denmark 600 240 1% -4 6

France 5,419 2,011 6% -32 53

Germany 39,734 16,916 51% -273 361

Greece 3,848 976 3% -18 21

Hungary 55 19 0% 0 0

Italy 19,691 7,168 21% -111 159

Luxembourg 30 12 0% 0 0

Macedonia

Montenegro

Netherlands 1,180 494 1% -8 11

Poland 7 2 0% 0 0

Portugal 322 53 0% -1 3

Romania 2,489 682 2% -14 16

Serbia 6 2 0% 0 0

Slovakia 525 188 1% -3 4

Slovenia 225 85 0% -1 2

Spain 6,739 1,392 4% -23 61

Switzerland 1,102 440 1% -7 10

Turkey 25 3 0% 0 0

Ukraine West n/a

Total 89,335 33,501 100%

Table 4 Installed capacity in MW, minute with the largest reduction in PV and the largest power gradient per minute due to PV

infeed per country based on the assumption of one geographical location per country.

Solar Eclipse 2015

10/21

Figure 4 Expected power gradient for the Top 5 countries based on the assumption of one geographical location per country.

5. Next Steps

Not all the TSOs will be affected by the eclipse on the same scale, but all will see the same impact on the

frequency; some countries are not affected by PV variations, but can support the other TSOs by providing

them reserves. The main challenge for the TSOs will be to coordinate the use of the reserves in order to

balance the power in real time without creating overloads on the grid. Therefore coordination procedures

should be exercised well in advance.

The proposed recommendations to the TSOs from Continental Europe synchronous region can be split in two

levels.

Individual TSOs

- The infeed from PV is highly depending on cloud coverage. The results presented in this report

assume clear sky conditions which may not appear. Day-ahead forecast of PV is particularly

important for March 20 and careful preparation and coordination of solar forecasts are necessary

between TSO-DSO-Balance responsible parties as well as between TSOs.

- Each TSO shall inform the Balance Responsible Parties (BRP) in charge of PV, and check that they

prepare the adequate measures to follow the PV variations on March, 20th, in order to remain balanced

on each program slot.

- Even if each BRP follows exactly the PV variation with intra-day programs, the TSOs will have to

balance the system within each program path slot, in order to follow the gradient of PV variations.

Each TSO shall estimate the amount of control reserves with adequate gradient (mainly Frequency

Restoration Reserve (whether automatic and manual)) will be needed for this specific situation.

- Each TSO shall increase control reserves as much as necessary for its own needs. In case of high

probability not to cover its own control block/area, each TSO shall estimate and declare to other

TSOs the amount of control reserve he will need in real time.

- TSOs which can provide more control reserve than they need shall propose these reserves to help

frequency management in real-time.

- Training of control room personnel for this specific event

Solar Eclipse 2015

11/21

- Due to change in thermostatic controlled loads and lighting, the demand could also be affected by

the solar eclipse. The effect is very difficult to estimate on a global scale. It is advised that each TSO

prepare its best estimate.

- Each TSO expecting significant changes in PV infeed shall estimate the available power regulation

speed and if possible check or test the capabilities.

TSOs expecting difficulties in balancing their control areas with their own reserves shall ensure the use of

reserves from other TSOs. Each TSO, in case of no possibility / high probability not to cover its own control

block/area, shall estimate the amount of reserves needed from other TSOs.

Continental Europe synchronous area coordination

- Impact on tie line flows cannot be properly assessed until a hypothesis for allocation of the balancing

power has been developed. However, before and during the eclipse the NTC on the critical borders

could be decreased in order to reserve as much capacity as needed for reserve exchanges.

- RSCIs1 can be involved in D-2, D-1 and ID to check forecast files, detect potential constraints due to

reserve exchanges, evaluate limits, propose remedial actions

- If necessary TSOs will set up an extraordinary operational coordination until the day of the eclipse,

including day-ahead and real-time teleconference to coordinate PV forecasts, real-time frequency

management, reserve exchanges and flow management.

1 A regional unified scheme set up by TSOs in order to coordinate Operational Security Analysis in a determined

geographic Area.

Solar Eclipse 2015

12/21

Analysis for the Great Britain Synchronous Area

1. Summary

Loss of Photo Voltaic (PV) infeed during the eclipse, and its return after maximum obscuration will occur

at a maximum rate of just below 50 MW/min.

The change in residual demand caused by human behaviour (halting normal activities to observe the

eclipse) will dominate the PV effect.

The PV effect acts in the opposite direction on the residual demand to the human-demand effect, and so

will in fact ameliorate the situation.

The rates of change of residual demand are therefore slightly less than experience during the previous

eclipse in 1999.

National Grid Transmission System coped well with the eclipse in 1999 as a result of careful planning.

We are confident that the system will cope well with the 2015 eclipse.

2. Model for Eclipse Effect

Data available

Start of eclipse tS

End of eclipse te

Maximum obscuration M

Result

At time t

% obscuration

= 2

𝜋 [ cos−1 √𝑑2 +

𝑎(𝑡)2

4⁄ − √𝑑2 +

𝑎(𝑡)2

4⁄ √1 − (𝑑2 +

𝑎(𝑡)2

4⁄ ) ]

where

𝑎(𝑡) = 𝐴 [2 (𝑡 − 𝑡𝑠

𝑡𝑒 − 𝑡𝑠) − 1]

𝐴 = 2√1 − 𝑑2 and d is evaluated by solving

𝑀 = 2

𝜋[ cos−1 𝑑 − 𝑑√1 − 𝑑2 ]

3. Model for the Residual Demand

We assume a bright day in March.

PV generation is assumed proportional to solar incident radiation. Solar radiation is obtained from UK

Metrological Office data at hourly intervals from 0800 to 1100 (GMT) on dates within 10 days of March 20

for all years in which we have data (since 2007). Maximum values are selected for each hour to represent a

clear day. We then interpolate with a smooth cubic spline to produce a smooth incident radiation curve, and

this is then transformed into a PV output.

The analysis has been done twice

Solar Eclipse 2015

13/21

First, with a single point approximation. Birmingham was chosen as being near centre of GB

system. Conditions for eclipse and solar radiation were found, and the model applied.

Secondly, GB was broken down into 28 regions, and the relevant conditions found for each region.

The results were then combined to provide a GB model.

As expected, the regional model reduces rates of change, and overall loss of PV, because of a smearing

effect as timing of eclipse is spread across the geographical region. However, the differences are relatively

small, so a single point approximation would be justified. Figure 1 shows PV generation during the eclipse

from the two models.

Figure 2 shows rates of change of PV generation for the two models.

Solar Eclipse 2015

14/21

The maximum rate of change caused by the variation in PV generation is approximately 50 MW/min. This

is well within the capabilities of the system.

Note that, as PV generation decreases, residual demand will increase, and vice versa. So rate of change of

demand as a consequence of PV will be positive before maximum eclipse, and negative afterwards.

4. Demand Effects

Based on evidence from the 1999 eclipse, we expect a depression in demand around the time of maximum

obscuration from human causes – people stopping work and going to look at the phenomenon…We refer to

this as the human-demand effect. This should depress demand before maximum eclipse – negative rate of

change, and increase demand after – positive rate of change.

These rates of change work in the opposite direction to the PV changes.

From 1999 evidence we expect the demand effect to dominate the PV effect.

5. Model for Residual Demand

A sample minute-by-minute demand curve is chosen to represent 20 March 2015. Data from 21 March

2014 is chosen. This demand curve is adjusted to account for the (modelled) PV generation for 21 March

2014 to give the underlying demand met by transmission system and PV generation.

The modelled PV generation with eclipse effects for a clear 20 March 2015 are then applied to give the

residual demand without the human-demand effect.

The human-demand effect is obtained by comparing demand as observed on the day of the 1999 eclipse

with a similar day from the same year (obviously without the eclipse).

The human-demand effect is applied to obtain a minute-by-minute illustrative demand curve for 20 March

2015.

Note: This is not a forecast, but an illustrative curve to give control engineers an idea of the type of curve

they will have to deal with.

Solar Eclipse 2015

15/21

The results are shown in Figure 3.

The greatest rates of change are caused by the human-demand effect rather than the PV effect, with a

maximum minutely rate of change of -300 MW/min, but typical values of -200 MW/min before maximum

eclipse, and +200 MW/min after.

The PV effect is most noticeable between 08:45 and 09:30 (all times are GMT), and 10:15 to 10:45. But

the rates of change here lie within +/- 150 MW/min. This rate of change is in a wider range than the PV

model on its own predicted because it is combined with the natural variability of the pure demand signal.

In 1999, with careful planning, the system coped with rates of change of demand comparable to this, and in

fact slightly higher, so National Grid is confident that this is a situation we have the experience and tools to

handle.

6. References

[1] “Global market outlook 2014,” EPIA,

http://www.epia.org/fileadmin/user_upload/Publications/EPIA_Global_Market_Outlook_for_Photovo

ltaics_2014-2018_-_Medium_Res.pdf, 2014.

[2] UK Hydrographic Office, [Online]. Available: http://astro.ukho.gov.uk/eclipse/0112015/.

[3] RTE Thierry Dellac, Simulation production éclipse 20-03-2015.xlsx.

Solar Eclipse 2015

16/21

Annex 1: Solar Obscuration Factor

Data available

Start of eclipse tS

End of eclipse te

Maximum obscuration M0

Result

At time t

% obscuration

= 2

𝜋 [ cos−1 √𝑑2 +

𝑎(𝑡)2

4⁄ − √𝑑2 +

𝑎(𝑡)2

4⁄ √1 − (𝑑2 +

𝑎(𝑡)2

4⁄ ) ]

where

𝑎(𝑡) = 𝐴 [2 (𝑡 − 𝑡𝑠

𝑡𝑒 − 𝑡𝑠) − 1]

𝐴 = 2√1 − 𝑑2 and d is evaluated by solving

𝑀𝑜 = 2

𝜋[ cos−1 𝑑 − 𝑑√1 − 𝑑2 ]

Assumptions

Assume :

Sun and Moon are the same size.

Centre of moon travels across Sun’s disc in a straight line at constant speed.

Proof Let

- radius of Sun and Moon = 1

d.1

Sun with Centre (Sc )

Solar Eclipse 2015

17/21

- distance of nearest approach of Moon’s centre = 2d

- as moon moves across Sun, let distance of Moon’s centre at time, t from closest approach to Sun be

a(t)

- distance of Moon’s centre from point of closest approach to Sun’s centre be A = a(ts) = a(te)

First consider instant of maximum obscuration

Calculate area of overlap of the two circles.

The two shaded areas are equal by symmetry

𝐴𝑛𝑔𝑙𝑒 𝜃 = 2 cos−1 𝑑

𝐴0 = 2𝜋 cos−1 𝑑

2𝜋 = cos−1 𝑑

Area of overlap

ϴ

Solar Eclipse 2015

18/21

= 2𝐴0 − 𝑎𝑟𝑒𝑎 𝑜𝑓 𝑟ℎ𝑜𝑚𝑏𝑢𝑠 = 2𝐴0 − 2𝑑√1 − 𝑑2 Overlap therefore

2 [cos−1 𝑑 − 𝑑√1 − 𝑑2]

% Overlap 2

𝜋 [cos−1 𝑑 − 𝑑√1 − 𝑑2 ] = 𝑀0

Given M0 we can calculate d [Note: this is not analytically invertible, but can easily be solved numerically]

Now consider obscuration when Moon’s centre is distance a(t) = a from point of closest approach

Variable area of overlap can be calculated in exactly the same way, but instead of separation of 2d, use the

separation between MC and SC

[ MC to SC ] = √𝑎2 + 4𝑑2 so replace d with √𝑎2

4+ 𝑑2

% Overlap at time t is

= 2

𝜋 [ cos−1 √𝑑2 +

𝑎(𝑡)2

4⁄ − √𝑑2 +

𝑎(𝑡)2

4⁄ √1 − (𝑑2 +

𝑎(𝑡)2

4⁄ ) ]

Now calculate distance from Moon’s centre to point of closest approach at the instant of first ( or last )

contact

𝐴 = 2√1 − 𝑑2 Finally find the distance of Moon’s centre from point of closest approach at time t. Moon moves from –A at

time, ts to +A at time te so a(t) = 𝐴 [2 (𝑡−𝑡𝑠

𝑡𝑒− 𝑡𝑠) − 1].

Putting this together we obtain stated result.

The formula is derived by Andrew Richards, National Grid, UK.

Solar Eclipse 2015

19/21

Annex 2: Solar eclipse data for each country

The eclipse start and end time and the maximum obscuration per country are derived from data calculated

by the UK Hydrographic Office made available on http://astro.ukho.gov.uk/eclipse/0112015/

Solar Eclipse 2015

20/21

Solar Eclipse 2015

21/21